Update app.py

app.py

@@ -70,7 +70,7 @@ class ChatBot:
         inputs = self.tokenizer(prompt, return_tensors="pt").to(self.model.device)
         
         # Set EOS seen in training (per Qwen chat template) 
-        stop_id = tokenizer.convert_tokens_to_ids("<|im_end|>")
+        stop_id = self.tokenizer.convert_tokens_to_ids("<|im_end|>")
         
         # Generate response
         with torch.no_grad():